diff --git a/dlls/wined3d/arb_program_shader.c b/dlls/wined3d/arb_program_shader.c
index ee56e9ce76b..a2b07dbe65d 100644
--- a/dlls/wined3d/arb_program_shader.c
+++ b/dlls/wined3d/arb_program_shader.c
@@ -144,6 +144,9 @@ void shader_generate_arb_declarations(
     unsigned max_constantsF = min(This->baseShader.limits.constant_float, 
             (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
 
+    /* Temporary Output register */
+    shader_addline(buffer, "TEMP TMP_OUT;\n");
+
     for(i = 0; i < This->baseShader.limits.temporary; i++) {
         if (reg_maps->temporary[i])
             shader_addline(buffer, "TEMP R%lu;\n", i);
@@ -168,6 +171,7 @@ void shader_generate_arb_declarations(
     /* Need to PARAM the environment parameters (constants) so we can use relative addressing */
     shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
                    max_constantsF, max_constantsF - 1);
+    shader_addline(buffer, "PARAM PROJECTION = state.matrix.projection.row[1];\n");
 }
 
 static const char* shift_tab[] = {
@@ -344,7 +348,7 @@ static void vshader_program_add_param(SHADER_OPCODE_ARG *arg, const DWORD param,
   IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) arg->shader;
 
   /* oPos, oFog and oPts in D3D */
-  static const char* hwrastout_reg_names[] = { "result.position", "TMP_FOG", "result.pointsize" };
+  static const char* hwrastout_reg_names[] = { "TMP_OUT", "TMP_FOG", "result.pointsize" };
 
   DWORD reg = param & D3DSP_REGNUM_MASK;
   DWORD regtype = shader_get_regtype(param);
diff --git a/dlls/wined3d/directx.c b/dlls/wined3d/directx.c
index 51d41e7af5d..5b3549418a8 100644
--- a/dlls/wined3d/directx.c
+++ b/dlls/wined3d/directx.c
@@ -247,8 +247,10 @@ void select_shader_max_constants(WineD3D_GL_Info *gl_info) {
             gl_info->max_vshader_constantsF = gl_info->vs_glsl_constantsF - MAX_CONST_B - MAX_CONST_I;
             break;
         case SHADER_ARB:
-            /* ARB shaders seem to have an implicit PARAM when fog is used, so we need to subtract 1 from the total available */
-            gl_info->max_vshader_constantsF = gl_info->vs_arb_constantsF - 1;
+            /* We have to subtract any other PARAMs that we might use in our shader programs.
+             * ATI seems to count 2 implicit PARAMs when we use fog and NVIDIA counts 1,
+             * and we reference one row of the PROJECTION matrix which counts as 1 PARAM. */
+            gl_info->max_vshader_constantsF = gl_info->vs_arb_constantsF - 3;
             break;
         case SHADER_SW:
             gl_info->max_vshader_constantsF = 96;  /* TODO: Fixup software shaders */
diff --git a/dlls/wined3d/drawprim.c b/dlls/wined3d/drawprim.c
index a7dd644d3ef..ab6162c14d5 100644
--- a/dlls/wined3d/drawprim.c
+++ b/dlls/wined3d/drawprim.c
@@ -304,9 +304,6 @@ static void primitiveInitState(
         }
 
         /* Vertex Shader output is already transformed, so set up identity matrices */
-        /* FIXME: Actually, only true for software emulated ones, so when h/w ones
-             come along this needs to take into account whether s/w ones were
-             requested or not                                                       */
         if (useVS) {
             glMatrixMode(GL_MODELVIEW);
             checkGLcall("glMatrixMode");
diff --git a/dlls/wined3d/vertexshader.c b/dlls/wined3d/vertexshader.c
index c0ee39957bb..923f953652d 100644
--- a/dlls/wined3d/vertexshader.c
+++ b/dlls/wined3d/vertexshader.c
@@ -718,6 +718,11 @@ static VOID IWineD3DVertexShaderImpl_GenerateShader(
             This->usesFog = 1;
             shader_addline(&buffer, "gl_FogFragCoord = clamp(gl_FogFragCoord, 0.0, 1.0);\n");
         }
+        
+        /* Write the final position.
+         * Account for any inverted textures (render to texture case) by reversing the y coordinate
+         *  (this is handled in drawPrim() when it sets the MODELVIEW and PROJECTION matrices) */
+        shader_addline(&buffer, "gl_Position.y = gl_Position.y * gl_ProjectionMatrix[1][1];\n");
 
         shader_addline(&buffer, "}\n\0");
 
@@ -754,6 +759,12 @@ static VOID IWineD3DVertexShaderImpl_GenerateShader(
         if (reg_maps->fog)
             shader_addline(&buffer, "MAX result.fogcoord, TMP_FOG, 0.0;\n");
 
+        /* Write the final position.
+         * Account for any inverted textures (render to texture case) by reversing the y coordinate
+         *  (this is handled in drawPrim() when it sets the MODELVIEW and PROJECTION matrices) */
+        shader_addline(&buffer, "MOV result.position, TMP_OUT;\n");
+        shader_addline(&buffer, "MUL result.position.y, TMP_OUT.y, state.matrix.projection.row[1].y;\n");
+        
         shader_addline(&buffer, "END\n\0"); 
 
         /* TODO: change to resource.glObjectHandle or something like that */